1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -scoped-noalias-aa -slp-vectorizer -mtriple=x86_64-apple-darwin -enable-new-pm=false -S %s | FileCheck %s
3; RUN: opt -aa-pipeline='basic-aa,scoped-noalias-aa' -passes=slp-vectorizer -mtriple=x86_64-apple-darwin -S %s | FileCheck %s
4
5define void @version_multiple(i32* nocapture %out_block, i32* nocapture readonly %counter) {
6; CHECK-LABEL: @version_multiple(
7; CHECK-NEXT:  entry:
8; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[COUNTER:%.*]], align 4
9; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[OUT_BLOCK:%.*]], align 4
10; CHECK-NEXT:    [[XOR:%.*]] = xor i32 [[TMP1]], [[TMP0]]
11; CHECK-NEXT:    store i32 [[XOR]], i32* [[OUT_BLOCK]], align 4
12; CHECK-NEXT:    [[ARRAYIDX_1:%.*]] = getelementptr inbounds i32, i32* [[COUNTER]], i64 1
13; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[ARRAYIDX_1]], align 4
14; CHECK-NEXT:    [[ARRAYIDX2_1:%.*]] = getelementptr inbounds i32, i32* [[OUT_BLOCK]], i64 1
15; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX2_1]], align 4
16; CHECK-NEXT:    [[XOR_1:%.*]] = xor i32 [[TMP3]], [[TMP2]]
17; CHECK-NEXT:    store i32 [[XOR_1]], i32* [[ARRAYIDX2_1]], align 4
18; CHECK-NEXT:    [[ARRAYIDX_2:%.*]] = getelementptr inbounds i32, i32* [[COUNTER]], i64 2
19; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX_2]], align 4
20; CHECK-NEXT:    [[ARRAYIDX2_2:%.*]] = getelementptr inbounds i32, i32* [[OUT_BLOCK]], i64 2
21; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[ARRAYIDX2_2]], align 4
22; CHECK-NEXT:    [[XOR_2:%.*]] = xor i32 [[TMP5]], [[TMP4]]
23; CHECK-NEXT:    store i32 [[XOR_2]], i32* [[ARRAYIDX2_2]], align 4
24; CHECK-NEXT:    [[ARRAYIDX_3:%.*]] = getelementptr inbounds i32, i32* [[COUNTER]], i64 3
25; CHECK-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX_3]], align 4
26; CHECK-NEXT:    [[ARRAYIDX2_3:%.*]] = getelementptr inbounds i32, i32* [[OUT_BLOCK]], i64 3
27; CHECK-NEXT:    [[TMP7:%.*]] = load i32, i32* [[ARRAYIDX2_3]], align 4
28; CHECK-NEXT:    [[XOR_3:%.*]] = xor i32 [[TMP7]], [[TMP6]]
29; CHECK-NEXT:    store i32 [[XOR_3]], i32* [[ARRAYIDX2_3]], align 4
30; CHECK-NEXT:    ret void
31;
32entry:
33  %0 = load i32, i32* %counter, align 4
34  %1 = load i32, i32* %out_block, align 4
35  %xor = xor i32 %1, %0
36  store i32 %xor, i32* %out_block, align 4
37  %arrayidx.1 = getelementptr inbounds i32, i32* %counter, i64 1
38  %2 = load i32, i32* %arrayidx.1, align 4
39  %arrayidx2.1 = getelementptr inbounds i32, i32* %out_block, i64 1
40  %3 = load i32, i32* %arrayidx2.1, align 4
41  %xor.1 = xor i32 %3, %2
42  store i32 %xor.1, i32* %arrayidx2.1, align 4
43  %arrayidx.2 = getelementptr inbounds i32, i32* %counter, i64 2
44  %4 = load i32, i32* %arrayidx.2, align 4
45  %arrayidx2.2 = getelementptr inbounds i32, i32* %out_block, i64 2
46  %5 = load i32, i32* %arrayidx2.2, align 4
47  %xor.2 = xor i32 %5, %4
48  store i32 %xor.2, i32* %arrayidx2.2, align 4
49  %arrayidx.3 = getelementptr inbounds i32, i32* %counter, i64 3
50  %6 = load i32, i32* %arrayidx.3, align 4
51  %arrayidx2.3 = getelementptr inbounds i32, i32* %out_block, i64 3
52  %7 = load i32, i32* %arrayidx2.3, align 4
53  %xor.3 = xor i32 %7, %6
54  store i32 %xor.3, i32* %arrayidx2.3, align 4
55  ret void
56}
57
58declare void @use(<8 x float>)
59define void @delete_pointer_bound(float* %a, float* %b, i1 %c) #0 {
60; CHECK-LABEL: @delete_pointer_bound(
61; CHECK-NEXT:  entry:
62; CHECK-NEXT:    [[B_10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 10
63; CHECK-NEXT:    [[B_14:%.*]] = getelementptr inbounds float, float* [[B]], i64 14
64; CHECK-NEXT:    br i1 [[C:%.*]], label [[THEN:%.*]], label [[ELSE:%.*]]
65; CHECK:       else:
66; CHECK-NEXT:    [[L0:%.*]] = load float, float* [[B_10]], align 4
67; CHECK-NEXT:    [[L1:%.*]] = load float, float* [[B_14]], align 4
68; CHECK-NEXT:    [[I2:%.*]] = insertelement <8 x float> undef, float [[L0]], i32 2
69; CHECK-NEXT:    [[I3:%.*]] = insertelement <8 x float> [[I2]], float [[L0]], i32 3
70; CHECK-NEXT:    [[I4:%.*]] = insertelement <8 x float> [[I3]], float [[L1]], i32 4
71; CHECK-NEXT:    [[I7:%.*]] = insertelement <8 x float> [[I4]], float [[L1]], i32 7
72; CHECK-NEXT:    call void @use(<8 x float> [[I7]])
73; CHECK-NEXT:    ret void
74; CHECK:       then:
75; CHECK-NEXT:    [[A_8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 8
76; CHECK-NEXT:    store float 0.000000e+00, float* [[A_8]], align 4
77; CHECK-NEXT:    [[L6:%.*]] = load float, float* [[B_14]], align 4
78; CHECK-NEXT:    [[A_5:%.*]] = getelementptr inbounds float, float* [[A]], i64 5
79; CHECK-NEXT:    store float [[L6]], float* [[A_5]], align 4
80; CHECK-NEXT:    [[A_6:%.*]] = getelementptr inbounds float, float* [[A]], i64 6
81; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A_6]] to <2 x float>*
82; CHECK-NEXT:    store <2 x float> zeroinitializer, <2 x float>* [[TMP0]], align 4
83; CHECK-NEXT:    ret void
84;
85entry:
86  %b.10 = getelementptr inbounds float, float* %b, i64 10
87  %b.14 = getelementptr inbounds float, float* %b, i64 14
88  br i1 %c, label %then, label %else
89
90else:
91  %l0 = load float, float* %b.10, align 4
92  %l1 = load float, float* %b.14, align 4
93  %i2 = insertelement <8 x float> undef, float %l0, i32 2
94  %i3 = insertelement <8 x float> %i2, float %l0, i32 3
95  %i4 = insertelement <8 x float> %i3, float %l1, i32 4
96  %i7 = insertelement <8 x float> %i4, float %l1, i32 7
97  call void @use(<8 x float> %i7)
98  ret void
99
100then:
101  %a.8 = getelementptr inbounds float, float* %a, i64 8
102  store float 0.0, float* %a.8, align 4
103  %l6 = load float, float* %b.14, align 4
104  %a.5 = getelementptr inbounds float, float* %a, i64 5
105  store float %l6, float* %a.5, align 4
106  %a.6 = getelementptr inbounds float, float* %a, i64 6
107  store float 0.0, float* %a.6, align 4
108  %a.7 = getelementptr inbounds float, float* %a, i64 7
109  store float 0.0, float* %a.7, align 4
110  ret void
111}
112
113%struct.zot = type { i16, i16, i16, i32, float, float, float, %struct.quux*, %struct.zot*, %struct.wombat*, %struct.wombat.0 }
114%struct.quux = type { i16, %struct.quux*, %struct.quux* }
115%struct.wombat = type { i32, i16, i8, i8, %struct.eggs* }
116%struct.eggs = type { float, i8, %struct.ham }
117%struct.ham = type { [2 x double], [8 x i8] }
118%struct.wombat.0 = type { %struct.bar }
119%struct.bar = type { [3 x double], [3 x double], double, double, i16, [3 x double]*, i32, [3 x double] }
120
121define double @preserve_loop_info(%struct.zot* %arg) {
122; CHECK-LABEL: @preserve_loop_info(
123; CHECK-NEXT:  entry:
124; CHECK-NEXT:    [[TMP:%.*]] = alloca [3 x double], align 16
125; CHECK-NEXT:    br label [[OUTER_HEADER:%.*]]
126; CHECK:       outer.header:
127; CHECK-NEXT:    br label [[INNER:%.*]]
128; CHECK:       inner:
129; CHECK-NEXT:    br i1 undef, label [[OUTER_LATCH:%.*]], label [[INNER]]
130; CHECK:       outer.latch:
131; CHECK-NEXT:    br i1 undef, label [[BB:%.*]], label [[OUTER_HEADER]]
132; CHECK:       bb:
133; CHECK-NEXT:    [[TMP5:%.*]] = load [3 x double]*, [3 x double]** undef, align 8
134; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[TMP]], i64 0, i64 0
135; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[TMP]], i64 0, i64 1
136; CHECK-NEXT:    br label [[LOOP_3HEADER:%.*]]
137; CHECK:       loop.3header:
138; CHECK-NEXT:    br i1 undef, label [[LOOP_3LATCH:%.*]], label [[BB9:%.*]]
139; CHECK:       bb9:
140; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[TMP5]], i64 undef, i64 1
141; CHECK-NEXT:    store double undef, double* [[TMP6]], align 16
142; CHECK-NEXT:    [[TMP12:%.*]] = load double, double* [[TMP10]], align 8
143; CHECK-NEXT:    store double [[TMP12]], double* [[TMP7]], align 8
144; CHECK-NEXT:    br label [[LOOP_3LATCH]]
145; CHECK:       loop.3latch:
146; CHECK-NEXT:    br i1 undef, label [[BB14:%.*]], label [[LOOP_3HEADER]]
147; CHECK:       bb14:
148; CHECK-NEXT:    [[TMP15:%.*]] = call double undef(double* [[TMP6]], %struct.zot* [[ARG:%.*]])
149; CHECK-NEXT:    ret double undef
150;
151entry:
152  %tmp = alloca [3 x double], align 16
153  br label %outer.header
154
155outer.header:                                              ; preds = %bb3, %bb
156  br label %inner
157
158inner:
159  br i1 undef, label %outer.latch, label %inner
160
161outer.latch:                                              ; preds = %bb16
162  br i1 undef, label %bb, label %outer.header
163
164bb:                                              ; preds = %bb3
165  %tmp5 = load [3 x double]*, [3 x double]** undef, align 8
166  %tmp6 = getelementptr inbounds [3 x double], [3 x double]* %tmp, i64 0, i64 0
167  %tmp7 = getelementptr inbounds [3 x double], [3 x double]* %tmp, i64 0, i64 1
168  br label %loop.3header
169
170loop.3header:                                              ; preds = %bb13, %bb4
171  br i1 undef, label %loop.3latch, label %bb9
172
173bb9:                                              ; preds = %bb8
174  %tmp10 = getelementptr inbounds [3 x double], [3 x double]* %tmp5, i64 undef, i64 1
175  store double undef, double* %tmp6, align 16
176  %tmp12 = load double, double* %tmp10, align 8
177  store double %tmp12, double* %tmp7, align 8
178  br label %loop.3latch
179
180loop.3latch:                                             ; preds = %bb11, %bb8
181  br i1 undef, label %bb14, label %loop.3header
182
183bb14:                                             ; preds = %bb13
184  %tmp15 = call double undef(double* %tmp6, %struct.zot* %arg)
185  ret double undef
186}
187
188define void @gather_sequence_crash(<2 x float> %arg, float* %arg1, float %arg2, float* %arg3, float* %arg4, float* %arg5, i1 %c.1, i1 %c.2) {
189; CHECK-LABEL: @gather_sequence_crash(
190; CHECK-NEXT:  bb:
191; CHECK-NEXT:    br i1 [[C_1:%.*]], label [[BB16:%.*]], label [[BB6:%.*]]
192; CHECK:       bb6:
193; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[ARG1:%.*]], i32 3
194; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x float> <float poison, float poison, float poison, float 0.000000e+00>, float [[ARG2:%.*]], i32 0
195; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x float> [[ARG:%.*]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
196; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> [[TMP1]], <4 x i32> <i32 0, i32 4, i32 5, i32 3>
197; CHECK-NEXT:    [[TMP3:%.*]] = fmul <4 x float> [[TMP2]], zeroinitializer
198; CHECK-NEXT:    [[TMP4:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
199; CHECK-NEXT:    store <4 x float> [[TMP3]], <4 x float>* [[TMP4]], align 4
200; CHECK-NEXT:    ret void
201; CHECK:       bb16:
202; CHECK-NEXT:    br label [[BB17:%.*]]
203; CHECK:       bb17:
204; CHECK-NEXT:    br label [[BB18:%.*]]
205; CHECK:       bb18:
206; CHECK-NEXT:    br label [[BB19:%.*]]
207; CHECK:       bb19:
208; CHECK-NEXT:    br label [[BB20:%.*]]
209; CHECK:       bb20:
210; CHECK-NEXT:    br label [[BB21:%.*]]
211; CHECK:       bb21:
212; CHECK-NEXT:    br label [[BB22:%.*]]
213; CHECK:       bb22:
214; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, float* [[ARG4:%.*]], i32 0
215; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i64 7
216; CHECK-NEXT:    br i1 [[C_2:%.*]], label [[BB25:%.*]], label [[BB22]]
217; CHECK:       bb25:
218; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr float, float* [[TMP23]], i64 6
219; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP24]], align 4
220; CHECK-NEXT:    [[TMP27:%.*]] = load float, float* [[ARG5:%.*]], align 4
221; CHECK-NEXT:    [[TMP29:%.*]] = fadd float 0.000000e+00, 0.000000e+00
222; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP26]], align 4
223; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr float, float* [[TMP23]], i64 4
224; CHECK-NEXT:    [[TMP31:%.*]] = fadd float 0.000000e+00, 0.000000e+00
225; CHECK-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP30]] to <2 x float>*
226; CHECK-NEXT:    store <2 x float> zeroinitializer, <2 x float>* [[TMP5]], align 4
227; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[ARG4]], i32 0
228; CHECK-NEXT:    br label [[BB33:%.*]]
229; CHECK:       bb33:
230; CHECK-NEXT:    br label [[BB34:%.*]]
231; CHECK:       bb34:
232; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr float, float* [[TMP32]], i64 3
233; CHECK-NEXT:    [[TMP37:%.*]] = load float, float* [[TMP35]], align 4
234; CHECK-NEXT:    [[TMP38:%.*]] = fadd float 0.000000e+00, [[TMP37]]
235; CHECK-NEXT:    store float [[TMP38]], float* [[TMP35]], align 4
236; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr float, float* [[TMP32]], i64 1
237; CHECK-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP39]] to <2 x float>*
238; CHECK-NEXT:    [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4
239; CHECK-NEXT:    [[TMP8:%.*]] = fadd <2 x float> zeroinitializer, [[TMP7]]
240; CHECK-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP39]] to <2 x float>*
241; CHECK-NEXT:    store <2 x float> [[TMP8]], <2 x float>* [[TMP9]], align 4
242; CHECK-NEXT:    [[TMP44:%.*]] = load float, float* [[ARG3:%.*]], align 4
243; CHECK-NEXT:    [[TMP45:%.*]] = load float, float* [[TMP32]], align 4
244; CHECK-NEXT:    [[TMP46:%.*]] = fadd float 0.000000e+00, [[TMP45]]
245; CHECK-NEXT:    store float [[TMP46]], float* [[TMP32]], align 4
246; CHECK-NEXT:    call void @quux()
247; CHECK-NEXT:    br label [[BB47:%.*]]
248; CHECK:       bb47:
249; CHECK-NEXT:    br label [[BB17]]
250;
251bb:
252  br i1 %c.1, label %bb16, label %bb6
253
254bb6:                                              ; preds = %bb
255  %tmp = getelementptr inbounds float, float* %arg1, i32 4
256  %tmp7 = getelementptr inbounds float, float* %arg1, i32 5
257  %tmp8 = getelementptr inbounds float, float* %arg1, i32 3
258  %tmp9 = getelementptr inbounds float, float* %arg1, i32 6
259  %tmp10 = extractelement <2 x float> %arg, i32 0
260  %tmp11 = fmul float %tmp10, 0.000000e+00
261  store float %tmp11, float* %tmp, align 4
262  %tmp12 = extractelement <2 x float> %arg, i32 1
263  %tmp13 = fmul float %tmp12, 0.000000e+00
264  store float %tmp13, float* %tmp7, align 4
265  %tmp14 = fmul float %arg2, 0.000000e+00
266  store float %tmp14, float* %tmp8, align 4
267  %tmp15 = fmul float 0.000000e+00, 0.000000e+00
268  store float %tmp15, float* %tmp9, align 4
269  ret void
270
271bb16:                                             ; preds = %bb
272  br label %bb17
273
274bb17:                                             ; preds = %bb47, %bb16
275  br label %bb18
276
277bb18:                                             ; preds = %bb17
278  br label %bb19
279
280bb19:                                             ; preds = %bb18
281  br label %bb20
282
283bb20:                                             ; preds = %bb19
284  br label %bb21
285
286bb21:                                             ; preds = %bb20
287  br label %bb22
288
289bb22:                                             ; preds = %bb22, %bb21
290  %tmp23 = getelementptr inbounds float, float* %arg4, i32 0
291  %tmp24 = getelementptr float, float* %tmp23, i64 7
292  br i1 %c.2, label %bb25, label %bb22
293
294bb25:                                             ; preds = %bb22
295  %tmp26 = getelementptr float, float* %tmp23, i64 6
296  store float 0.000000e+00, float* %tmp24, align 4
297  %tmp27 = load float, float* %arg5, align 4
298  %tmp28 = getelementptr float, float* %tmp23, i64 5
299  %tmp29 = fadd float 0.000000e+00, 0.000000e+00
300  store float 0.000000e+00, float* %tmp26, align 4
301  %tmp30 = getelementptr float, float* %tmp23, i64 4
302  store float 0.000000e+00, float* %tmp28, align 4
303  %tmp31 = fadd float 0.000000e+00, 0.000000e+00
304  store float 0.000000e+00, float* %tmp30, align 4
305  %tmp32 = getelementptr inbounds float, float* %arg4, i32 0
306  br label %bb33
307
308bb33:                                             ; preds = %bb25
309  br label %bb34
310
311bb34:                                             ; preds = %bb33
312  %tmp35 = getelementptr float, float* %tmp32, i64 3
313  %tmp36 = getelementptr float, float* %tmp32, i64 2
314  %tmp37 = load float, float* %tmp35, align 4
315  %tmp38 = fadd float 0.000000e+00, %tmp37
316  store float %tmp38, float* %tmp35, align 4
317  %tmp39 = getelementptr float, float* %tmp32, i64 1
318  %tmp40 = load float, float* %tmp36, align 4
319  %tmp41 = fadd float 0.000000e+00, %tmp40
320  store float %tmp41, float* %tmp36, align 4
321  %tmp42 = load float, float* %tmp39, align 4
322  %tmp43 = fadd float 0.000000e+00, %tmp42
323  store float %tmp43, float* %tmp39, align 4
324  %tmp44 = load float, float* %arg3, align 4
325  %tmp45 = load float, float* %tmp32, align 4
326  %tmp46 = fadd float 0.000000e+00, %tmp45
327  store float %tmp46, float* %tmp32, align 4
328  call void @quux()
329  br label %bb47
330
331bb47:                                             ; preds = %bb34
332  br label %bb17
333}
334
335declare void @quux()
336attributes #0 = { "target-features"="+avx2" }
337